AndrewZhaoLuo commented on a change in pull request #8069:
URL: https://github.com/apache/tvm/pull/8069#discussion_r652134539
##########
File path: python/tvm/relay/transform/transform.py
##########
@@ -1199,3 +1198,22 @@ def FakeQuantizationToInteger():
The registered SimplifyExpr pass.
"""
return _ffi_api.FakeQuantizationToInteger()
+
+
+def ToMixedPrecision(
+ mixed_precision_type="float16", ignore_missing_ops=True,
warn_missing_ops=True
+):
+ """
+ Automatic mixed precision rewriter. Rewrite an FP32 relay graph into a
version
+ where as many operations as possible are in the target
mixed_precision_type.
+
+ Note this does mutate the original graph putting it in a bad state
potentially.
+
+ TODO(AndrewZhaoLuo): don't mutate the original graph.
Review comment:
That problem is pretty old. It doesn't seem to have the problem anymore
so removed.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
